Depends on whether you want “exactly” what the user typed or the systems interpretation.  For the system’s interpretation, merely do:
u: (_*) $$usertyped = ^join( $$usertyped ” ” ‘_0) 
which will accumulate it and later you can use
u: (!%more) 
to know there are no more input sentences and do something with $$usertyped

If you want LITERALLY what they typed, the easiest way is to have all your tokenflags turned off except for system essentials (so no spell checking, no substitutions, etc) although you still get the system’s tokenization processing.  Memorize that, then turn on the normal flags and do a retry on the input so you get normal processing as well. Then you have to turn off tokenflags at the end, ready for the next round.  (one can reverse the order of whether tokenflags are on or off, but its more cumbersome with on first, since you have to deal with the output generated).

Thanks for the quick response! I’ll play with those and see what works.

I’m trying to create a seperate log file for just those inputs that the bot has to respond “I don’t know” to. If I can log what the user typed, then I only have to look at that single log file to see what the bot needs to “learn”.

then the
u: (_*)  $$usertyped = ‘_0
should be sufficient since you care about don’t knows on a single input at a time, not the collection of a volley

Thanks Bruce. That’s working nicely.

If anyone else is interested:

In simplecontrol.top:

Near line 21 I changed:
# main per-sentence processing
u: ()
to u: (_*)  $$usertyped = ‘_0

and added

^log(FILE TMP/noanswer.txt $$usertyped \n)

around line 104, just before it finishes the final loop and says “I don’t know”.

This will give you a log file called noanswer.txt in your TMP directory. It will list, one per line, every user input that exhausts the simplecontrol.top script and forces the bot to respond with “I don’t know”. You can then use this information to create or update your topics, gambits, KB, etc.

Well that seems pretty handy. I will have to try that.

Nice to hear from you again, Doug. I’m glad someone else can use this. I’m working on some other custom functions, and also the PostgreSQL functionality. So far, it’s working beautifully. I have a 38gig database, and it can pull out records and list them in no time at all.

38 gig?  What have you (or rather your database) been ingesting?

LOL…I’ve loaded several ontologies. This isn’t a “working” database. I just wanted a very large database to test the loading and response times.  As a side note, it took almost 34 hours just to index the database, but I’m searching for terms within records and pulling out lists from within CS in just a couple of milliseconds.

I’m testing an enterprise-level database because I see *HUGE* enterprise-level potential for CS since you’ve added the database integration. I’m testing it under heavily loaded conditions so I can answer those questions as they come up.

Take the Virtual Patient System, for example. They could load a medical ontology and the virtual patient could have any number of symptoms or diseases for the student to try to identify. And my list of other ideas for it goes on and on.

My business background has been almost exclusively doing data networking and project management for Fortune 5 (not 500) corporations. When I see what your database integration is capable of…..it makes me want to throw some demos together and dig out my phonebook. lol

Off Topic (just for a second):

Doug… Possibly some here may have done medical A.I. stuff too **
May even be published in some textbook by some university somewhere. 

** Hint: Virtual Patient, meet your forerunner… Patient Simulator.

I’ve also done some work on a virtual patient project, but it was all under an NDA, so I cannot discuss details, or even generalities for that matter, for another 4 years.